LLM cost dashboard shows zero tokens error tracking fails

Not specified by the original author.

TL;DR

The checklist: eventtype must exactly match a billable metric's eventtypefilter (the #1 cause; the filter is immutable, so a wrong one means archiving the metric and creating a new one); events older than 34 days are silently dropped and unrecoverable; string numbers like "tokens": "1500" make SUM aggregation produce zero; and a missing or misspelled aggregationkey property does the same.

Fix

1. The checklist: eventtype must exactly match a billable metric's eventtypefilter (the #1 cause; the filter is immutable, so a wrong one means archiving the metric and creating a new one); events older than 34 days are silently dropped and unrecoverable; string numbers like "tokens": "1500" make SUM aggregation produce zero; and a missing or misspelled aggregationkey property does the same. Expected: you can confirm this matches your situation.

When to use

• You are setting up or using this Metronome feature.
• The symptom matches: sending usage events but invoice shows zero usage - silent drops.

When NOT to use

• Unrelated Metronome issues (different feature, different failure).
• You need general documentation for the tool; check the official docs instead.

Compatibility

Reported against Metronome. Source: https://github.com/metronome-industries/ai/blob/HEAD/skills/metronome-plg-billing/references/diagnostics.md.

Variant phrasings

sending usage events but invoice shows zero usage - silent drops

Why it happens

Usage events are ingested by Metronome but the invoice shows zero usage. Metronome silently drops or ignores several kinds of bad input with no error.

Edge cases

• If your error message differs even slightly, this is probably a different issue; search the exact text.
• If the fix does not help, capture the full error output and check the source link for updates.

Source

https://github.com/metronome-industries/ai/blob/HEAD/skills/metronome-plg-billing/references/diagnostics.md

Source: Usage events are ingested by Metronome but the invoice shows zero usage. Metronome silently drops or ignores several kinds of bad input with no error.

Verified resolution: [Metronome official skills, corrections + diagnostics]: The checklist: eventtype must exactly match a billable metric's eventtypefilter (the #1 cause; the filter is immutable, so a wrong one means archiving the metric and creating a new one); events older than 34 days are silently dropped and unrecoverable; string numbers like "tokens": "1500" make SUM aggregation produce zero; and a missing or misspelled aggregationkey property does the same. Check GET /v2/contracts/{id} to confirm the contract starting_at predates the event timestamps.

Source URL: https://github.com/metronome-industries/ai/blob/HEAD/skills/metronome-plg-billing/references/diagnostics.md

Not specified by the original author.

TL;DR: Your request logs say 2k tokens but the bill says 9k. The gap is tokens you never see: hidden reasoning tokens, image tokens, and cache writes. Confirm the cause Check the usage breakdown on a real response: 1. On reasoning models, usage.outputtokensdetails.reasoning_tokens counts internal thinking: billed as output, never returned in the response.

The fix

TL;DR: Your request logs say 2k tokens but the bill says 9k. The gap is tokens you never see: hidden reasoning tokens, image tokens, and cache writes. Confirm the cause Check the usage breakdown on a real response: 1. On reasoning models, usage.outputtokensdetails.reasoning_tokens counts internal thinking: billed as output, never returned in the response.

The symptom

Billed tokens consistently exceed what your prompt plus visible response suggest. The usage object has fields your logging ignores.

Confirm the cause

Check the usage breakdown on a real response:

1. Hidden reasoning tokens. On reasoning models, usage.output_tokens_details.reasoning_tokens counts internal thinking: billed as output, never returned in the response. A "short" answer can carry thousands of reasoning tokens. This is the biggest surprise on o-series and GPT-5.
2. Image tokens. Vision inputs are tokenized by size and detail level and land in input tokens. A few high-detail images can outweigh a long text prompt.
3. Tool and schema tokens. Tool definitions and structured output schemas count as input tokens on every call.
4. Cache write tokens. On GPT-5.6 and later, prompt_tokens_details.cache_write_tokens reports prompt tokens written to cache, billed at 1.25x the uncached input rate. Cache writes are not free on newer models.
5. Retries and streams. Every retried request rebills its prompt tokens. Every retried stream rebills them too.

The fix

• Log the full usage object, not just prompt and completion totals. Break down reasoningtokens, cachedtokens, and cachewritetokens per call.
• Budget reasoning models with reasoning in mind: lower reasoning_effort when deep thinking is not needed, and size max_completion_tokens for reasoning plus visible output.
• Right-size image detail. low instead of high when the task allows it.
• Fix retry storms and loops first; they multiply every hidden cost above.

Verify the fix

Compare your logged token totals against the billed totals for a day. After the fix they should reconcile within a few percent. Any remaining gap has a named field in the usage object; find it before it finds your budget.

When to use this

• This covers exactly what the title says: OpenAI billed more tokens than your logs show.
• You are setting this up for the first time, or auditing an existing setup.
• You want the key gotchas in one place before you start.

When not to use this

• You are doing a different workflow with OpenAI; these steps are specific to the title above.
• You need the full reference docs; this is the short path, not the manual.

Compatibility

• Mentions 5.6 in the original; behavior may differ on much older or newer releases.

OpenAI's reasoning guide documents reasoningtokens under usage outputtokensdetails: internal reasoning billed as output but never returned. The prompt caching guide documents cachewrite_tokens on GPT-5.6 and later, billed at 1.25x the uncached input rate, and confirms images, tools, and schemas count as input tokens (https://platform.openai.com/docs/guides/reasoning, https://platform.openai.com/docs/guides/prompt-caching).

Not specified by the original author.

TL;DR

Set a budget or spending limit per project in the dashboard before launch. Keys used in browsers or client apps must be ephemeral session keys minted server-side, never long-lived secret keys shipped in a bundle. Look for: token growth without traffic growth, a model upgrade that quietly doubled cost per request, batch jobs that should have been cached prompts.

When to use

You are seeing this: Split work into projects: one project per environment (dev, staging, prod) and per major workload. Use this skill when you run into "OpenAI cost control: project budgets, token limits, key scoping, and alerts".

When not to use

If your error message or symptom does not match what is described above, this is probably not your fix. Search for your exact error text instead of forcing this one to fit.

Versions

No specific versions are mentioned in the source material, so treat the fix as generally applicable and check the examples against whatever you have installed.

Cost control, layer by layer

The steps

1. Split work into projects: one project per environment (dev, staging, prod) and per major workload. Project-scoped API keys mean a leaked dev key cannot touch production spend.
2. Set a budget or spending limit per project in the dashboard before launch. The alert that fires at 80% of budget is worth more than the invoice review at 200%.
3. Watch the rate-limit headers, especially x-ratelimit-limit-project-tokens. Project-level token limits cap how fast one project can burn; treat a project hitting its limit as a signal to investigate, not just a limit to raise.
4. Scope every key to the minimum it needs. Keys used in browsers or client apps must be ephemeral session keys minted server-side, never long-lived secret keys shipped in a bundle.
5. Log token usage per feature, not just per project. Attribute spend to the feature that caused it: embeddings, batch jobs, realtime sessions, evals. The line item that surprises you is the one you were not tracking.
6. Review weekly for the first month, monthly after. Look for: token growth without traffic growth, a model upgrade that quietly doubled cost per request, batch jobs that should have been cached prompts.

The trap

One org-wide key and no budgets. Everything works until a bug loops on a reasoning model overnight. By the time finance notices, the money is gone and nobody knows which feature spent it.

Checklist

• Separate projects and keys per environment.
• Budgets or limits set before launch, alerts firing.
• Project token limits understood and monitored.
• No long-lived keys in client-side code.
• Spend attributed per feature and reviewed on a schedule.

OpenAI's rate limits guide documents x-ratelimit-limit-project-tokens and the per-header limit fields. Verified live at https://developers.openai.com/api/docs/guides/rate-limits.